ctrl: nro: reconcile on node group with poolName

Node group can be set with either MCPselector or a poolName. Validation
of the nodegroup configuraion is checked in the reconcile logic, and
because poolName is set and MCPSelector is not a valid nodeGroup we need
to consider this and reconcile.

This was missed when nodeGroup.PoolName was first introduced and didn't
cause any observed reconcile failures because:
1. any change on the nodeGroups would anyway trigger reconciliation
2. so far the MCP related updates were caused by Kubeletconfig updates
   which wasn't blocked until the paused-MCP case got handled (which is
why this appeared now).

Signed-off-by: Shereen Haj <shajmakh@redhat.com>

Author: shajmakh

internal/controller/numaresourcesoperator_controller.go

@@ -663,6 +663,18 @@ func (r *NUMAResourcesOperatorReconciler) mcpToNUMAResourceOperator(ctx context.
 		nro := &nros.Items[i]
 		mcpLabels := labels.Set(mcp.Labels)
 		for _, nodeGroup := range nro.Spec.NodeGroups {
+			if nodeGroup.PoolName != nil {
+				if mcp.Name == *nodeGroup.PoolName {
+					requests = append(requests, reconcile.Request{
+						NamespacedName: client.ObjectKey{
+							Name: nro.Name,
+						},
+					})
+					break
+				}
+				continue
+			}
+
 			if nodeGroup.MachineConfigPoolSelector == nil {
 				continue
 			}
@@ -679,11 +691,13 @@ func (r *NUMAResourcesOperatorReconciler) mcpToNUMAResourceOperator(ctx context.
 						Name: nro.Name,
 					},
 				})
+				break
 			}
 		}
 	}
 
 	return requests
+
 }
 
 func (r *NUMAResourcesOperatorReconciler) configMapToNUMAResourceOperator(ctx context.Context, cmObj client.Object) []reconcile.Request {